How to reproduce 3D synthesis example from 4O press release?

I tried to recreate the 3D synthesis example from the press release (in the “Select sample” list box, select “3D object synthesis”). In the press release, the output they got is a simple untextured model well suited for 3D reconstruction, but if I use the same prompts, I get more complex responses with wrong logo, wrong text and different models used for “view 0” and “view 1”, etc. Is it possible to recreate the example using ChatGPT or using the API?


I’m trying it too and it just keeps failing. I’m wondering if this is a specific functionality being rolled out over time.